Anthropic Turned a Profit and Wall Street Sold Off Anyway

On Wednesday, Anthropic walked into a closed-door meeting in San Francisco carrying the best numbers in its short life: $16.8 billion in quarterly revenue, up 45 percent, and — for the first time ever — a profit. Then the reports leaked out, and Wall Street did the unthinkable. It sold.

CoreWeave collapsed 16 percent. Dell gave up 11 percent. Oracle dropped 9 percent. Some $80 billion of market value evaporated in a single session while the company at the centre of it all was busy telling investors it had never made more money. If that sounds backwards, it isn't. It's the most honest day the AI trade has had all year — because what Anthropic really told investors wasn't that the business works. It's that the business works while the bills get bigger.

We cross-checked the Reuters exclusive, the FT follow-up and the Investor's Business Daily market recap against each other while writing this piece. The numbers match to the decimal. What doesn't match is the story each version tells about what "profitable" actually means.

The numbers: $16.8 billion and the word "adjusted"

The figures come from a pre-IPO investor meeting Anthropic held with company leadership in San Francisco on Wednesday, and they were reported first by Reuters, citing an investor who attended. Third-quarter revenue rose 45 percent to $16.8 billion from the second quarter. The first quarter had been $4.7 billion; the second $11.6 billion. Annualise the third quarter and you get $67.2 billion — a number that would make Anthropic, overnight, one of the largest software businesses on the planet.

Then came the kicker. The Financial Times reported on Wednesday evening that Anthropic had also told investors it turned a profit in the quarter — the first in company history. The FT's dispatch hit the wire at 22:14 UTC, which is 6:14 in the evening in New York, after the closing bell. That is the hour you choose when you want traders to sleep on your news.

But there was an asterisk, and the asterisk did the real talking. Investor's Business Daily, summarising the reports, noted the company described itself as profitable on an "adjusted basis." Adjusted for what, exactly? Anthropic hasn't said — it declined to comment on any of the figures, as it has all year — but anyone who has read the company's financing paperwork can guess. You don't adjust away costs unless the costs are the story.

The context for Wednesday's briefing was set a day earlier, when Anthropic let prospective investors interrogate its executives in San Francisco ahead of its planned listing — an event we covered in detail in our report on Anthropic's pre-IPO investor grilling. That session was about the pitch. Wednesday was about the proof. And the proof came with fine print.

San Francisco's Financial District, where Anthropic briefed institutional investors on its third-quarter results on Wednesday, October 14. Photo: Dale Cruse via Wikimedia Commons (CC BY 2.0).

Wednesday's bloodbath: good news, meet a hot CPI

To be fair to the market, Anthropic's numbers didn't land in a vacuum. They landed on top of a red-hot core inflation report that made the Federal Reserve's next move look hawkish again, and inflation is poison for a sector that borrows the way AI does. Major AI stocks slumped across the board, with the S&P 500 and Nasdaq suffering sharp losses and testing critical support levels, according to Reuters and Investor's Business Daily.

The casualty list was brutal. CoreWeave, the neocloud poster child, fell 16 percent while the yield on its 9.625% bonds due in 2032 screamed past 14 percent — the bond market charging the company ever more for the privilege of lending it money. Nebius dropped 11 percent. Oracle fell 9 percent. Dell plunged 11 percent, breaking below a key support level, a day after Barclays downgraded it and two other enterprise hardware names. AMD and Snowflake tumbled below their buy points. After the close, S&P 500 futures were down 0.3 percent and Nasdaq-100 futures down 0.6 percent, with AI names extending their losses on the Anthropic news.

The human texture of the day came from the credit desk, not the equity floor. "The era of easy money, which we experienced in the first nine months of the year, is starting to disappear and dissipate," Luis Alvarado, co-head of global fixed income strategy at Wells Fargo Investment Institute, told reporters. "Now the extra dollar or the marginal dollar is going to be more looked at with a microscope."

And then there was the quote that explained the whole session. "We're getting a whole lot of information suggesting that the largest AI companies may not be generating the earnings or revenue people anticipated, and I think that makes everyone nervous about companies like CoreWeave," Kim Forrest, chief investment officer at Bokeh Capital Partners, told Reuters. CoreWeave CEO Mike Intrator disagreed, of course — he wrote in an internal memo that the market was acting "irrationally," Bloomberg reported. When the CEO of the company down 16 percent calls the market irrational, the market usually has receipts.

The Hock Tan problem: capital, not chips

Here is the uncomfortable frame that explains why $16.8 billion and a profit can spook a market instead of soothing it. On Monday, two days before Anthropic's briefing, Broadcom CEO Hock Tan stood on a stage in Singapore at the DBS Asia Future Forward Conference and named the binding constraint on the entire AI build-out. It wasn't wafers. It wasn't memory. It wasn't power.

"The thing you hit on that I think might be the biggest constraint is capital," Tan said, answering DBS group chief executive Tan Su Shan, who had raised the rising cost of debt for the hyperscalers. He added that if Anthropic and OpenAI kept earning what they earn now from frontier models and products like coding and personal assistants, the cost of capital wouldn't be a problem: "They can afford to pay whatever coupons you like to charge."

That conditional — if they keep earning — is doing all the work. Intel CEO Lip-Bu Tan, on the same stage, put it even more bluntly: "If there's any challenge in the front, it's capital and the cost of capital, and then the other part is in the longer run, people want to see the ROI." Two chip chief executives, in the same week, warning that the industry's scarcest resource is no longer silicon. It's money. We examined that warning at length in our coverage of Hock Tan's capital-over-chips thesis, and Wednesday's tape was that thesis playing out in real time.

Because here's the arithmetic Wall Street was doing while the confetti was still in the air. Anthropic's $16.8 billion quarter annualises to $67.2 billion. Impressive — until you stack it against the company's commitments. Reuters reported last month that Anthropic has committed hundreds of billions of dollars in future spending under long-term agreements for cloud services, computing and infrastructure. Reuters also reported earlier this month, citing the company's IPO filing, that Broadcom has agreed to lend Anthropic up to $42 billion — the chipmaker literally financing its own customer's chip purchases. When your supplier has to lend you the money to buy its products, "adjusted basis" profitability starts to look less like a milestone and more like a mood.

Rows of server racks in a data centre — the kind of compute capacity Anthropic has committed hundreds of billions of dollars to secure under long-term agreements. Photo: Fleshas via Wikimedia Commons (CC BY-SA 3.0).

Anthropic vs. OpenAI: dueling disclosures

The market's nerves were already frayed before Wednesday, because last week OpenAI did something similar and it went badly. During a fundraising process, OpenAI shared revised revenue figures of $50 billion with existing investors — well below the $70 billion figure that had been widely known in the market. A person familiar with the matter explained that the gap reflected the exclusion of cloud-partner revenue, not a collapse in demand, according to market reports. The explanation didn't land. Investor's Business Daily noted this week that "AI fears revived on OpenAI revenue last week" — the confusion over accounting treatment put semiconductor shares and the Nasdaq under pressure then, too.

That episode set the trap for Anthropic. When one frontier lab's numbers need a glossary to be understood, the next lab's numbers arrive pre-suspect. And Anthropic's history gives sceptics plenty to work with: Reuters' earlier read of the company's confidential prospectus put 2025 revenue at about $4.6 billion against roughly $42 billion in net losses — with about $34 billion of that an accounting charge and operating losses near $8 billion. A company can grow revenue from $4.7 billion to $16.8 billion in three quarters and still be, in Reuters' own description of the IPO pitch, a "heavily unprofitable AI developer pursuing massive scale." Both things are true at once. That's the whole problem.

One more disclosure question now hangs over the IPO. The timing is no longer speculative: Reuters reports the public-market debut is likely to wait until after the November US midterm elections, and the company is targeting a valuation of more than $2 trillion — which would make it an early test of public-market appetite for the entire AI-at-scale trade before rival OpenAI lists. A listing that size, carrying accounting this creative and commitments this large, is either the trade of the decade or the moment the music stops. There isn't much middle ground.

Our read: the profit that wasn't, and the selloff that was right

Let's call this what it is. Anthropic did not prove on Wednesday that the AI business model works. It proved that a company can sell an astonishing amount of software while contractually owing creditors sums that dwarf what it sells. "Adjusted basis" profitability is corporate cosmetics — it is what you report when the unadjusted number would start a conversation you'd rather not have. The market, to its credit, refused to be flattered by it.

The colossally misunderstood part of this story is the idea that revenue was ever the question. Nobody seriously doubted Anthropic could sell Claude — the demand has been obvious for years. The question was always the one Hock Tan named in Singapore: capital. Can the revenue outrun the hundreds of billions already promised to cloud providers, chip designers and landlords before the lenders lose their nerve? Wednesday's answer was a shrug from the bond market and a 16 percent haircut for the sector's most leveraged player. The shoddy part is how long the industry has pretended the financing was a footnote instead of the thesis.

There is a genuine bull case here, and it deserves one sentence: $4.7 billion to $16.8 billion in three quarters is a growth rate most public companies never touch in their lifetimes, and if the compute commitments were sized against revenue that keeps compounding, the leverage works in Anthropic's favour. But leverage cuts both ways, and the market just reminded everyone which way is faster. Whether Anthropic can keep growing fast enough to outrun its own bills — not chips, not models, but bills — will decide whether this IPO is remembered as the AI era's coming-out party or its last good quarter. We'd bet the bond market has a view.

For the wider context on how the industry financed this build-out, see our earlier report on AI's $113 billion debt machine hitting the wall — Wednesday's rout was that story's next chapter.

Sources

Reporting: Reuters exclusives on Anthropic's third-quarter revenue (October 14, 2026) and the Financial Times' profit report via Reuters (October 14, 2026, 22:14 UTC); Investor's Business Daily market coverage ("Dow Jones Futures: Dell, AMD, CoreWeave Dive In AI-Led Sell-Off," October 14/15); Reuters on the AI selloff and hot CPI report (October 14); Reuters on Broadcom CEO Hock Tan's capital warning at the DBS Asia Future Forward Conference, Singapore (October 12); Computer Weekly's account of the Hock Tan session; Invezz on the CoreWeave selloff (October 14); market reports on OpenAI's revised revenue disclosure; the Motley Fool's account of Reuters' read of Anthropic's confidential prospectus. Anthropic declined to comment on the figures, as reported by CNBC and Reuters.

By TechPulse USA — TechPulse USA is an independent tech news publication covering the global AI race with a US-first lens. We verify every fact we print, link our sources, and separate what is proven from what is claimed.